N-GLYCOSYLATION OF CD97 WITHIN THE EGF DOMAINS IS CRUCIAL FOR
EPITOPE ACCESSIBILITY IN NORMAL AND MALIGNANT CELLS AS WELL
AS CD55 LIGAND BINDING
Manja WOBUS1, Bjo ¨rn VOGEL2, Eike SCHM¨ UCKING1, Jo ¨rg HAMANN3and Gabriela AUST1*
1Institute of Anatomy, University of Leipzig, Leipzig, Germany
2Department of Conservative Dentistry and Periodontology, University of Leipzig, Leipzig, Germany
3Laboratory for Experimental Immunology, Academic Medical Center, University of Amsterdam, Amsterdam, The Netherlands
CD97 is an EGF-TM7 receptor found on various carcino-
mas where expression levels correlate with dedifferentiation
and tumor stage, smooth muscle cells and leukocytes. CD97
acts as an adhesion molecule by binding to its cellular ligand,
CD55. In this study, we demonstrate that 2 immunodomi-
nant CD97 epitopes are not equally present in the various
cell types. Differences were apparent in gastrointestinal tu-
mors and smooth muscle cells where monoclonal antibodies
(mAbs) to the first epidermal growth factor (EGF) domain
(CD97EGF) showed a more restricted staining pattern than
mAbs to the stalk region (CD97stalk). This discrepancy was
not detectable in cultured gastrointestinal tumor cell lines. In
fact, the selection of the CD97 mAb influences the result of
clinical studies. Thus, we clarified the reason(s) for these
differences in CD97 mAb staining on various cell types. We
provide evidence that epitope accessibility for CD97EGF
mAbs depends on N-glycosylation. Immunoprecipitation of
CD97 from the Colo 205 tumor cell line revealed the estab-
lished 78 and 83 kDa products, while a 52 and 57 kDa band
were obtained from smooth muscle cells. N-glycosidase F
reduced the size of CD97 in Colo 205 cells to 52–57 kDa.
Culturing these cells with tunicamycin resulted in the same
decrease in size and impaired CD97EGFmAb binding. As
shown by site-directed mutagenesis, deletion of the N-glyco-
sylation sites located within the EGF domains efficiently dis-
turbed CD97EGFmAb immunoreactivity and, importantly,
binding of CD55. In conclusion, CD97EGFepitope accessibility
for mAbs and ligand binding is influenced by cell type-specific
© 2004 Wiley-Liss, Inc.
Key words: CD97; colorectal cancer; smooth muscle cell; epitope
CD97 is a member of a subfamily of 7-span transmembrane
(TM7) receptors referred to as EGF-TM7 proteins.1,2In normal
tissues, CD97 is abundantly expressed in smooth muscle cells,
macrophages, granulocytes, dendritic cells and in some T and B
cells.3Carcinomas of different origin showed higher CD97 expres-
sion compared to the corresponding normal cell types.4–7
CD97 consists of an extended extracellular region with several
N-terminal epidermal growth factor (EGF) domains coupled to the
TM7 domain by a stalk.1,8Various CD97 isoforms exist possess-
ing 3 (EGF 1,2,5), 4 (EGF 1,2,3,5), or 5 (EGF 1,2,3,4,5) EGF
domains. CD97 binds to membrane receptor CD55.9This interac-
tion is mediated by the first 2 CD97 EGF domains.9–11Chondroitin
sulfate glycosaminoglycan, present in cell membranes and the
extracellular matrix, is a new ligand to the longest CD97 isoform
The ability of CD97 EGF domains to interact with cellular
and/or extracellular matrix ligands suggests that the molecule is
involved in cell-cell or cell-matrix interactions. Various lines of
evidence have shown an important role for CD97 in tumor dedi-
fferentiation, migration, invasiveness and metastasis. CD97 levels
correlate with the in vitro migration and invasion capacity of
colorectal tumor cell lines and CD97-transfected cells.6Colorectal
carcinomas with more strongly CD97-stained, scattered tumor
cells at the invasion front showed a poorer clinical stage as well as
increased lymph vessel invasion compared to cases with uniform
CD97 staining.6In thyroid cancers, CD97 expression parallels the
aggressiveness and lymph node involvement of these tumors.4,5
Our previous data revealed that CD97 has prognostic and ther-
apeutic potential in cancer. However, analyzing CD97 expression
brings up the phenomenon that different staining patterns, depend-
ing on which CD97 monoclonal antibody (mAb) has been used for
immunohistology, can be readily observed in pancreatic tissues.7,13
Based on this observation, one aim of the present study was to
examine in detail the differences between several CD97 mAbs in
normal and malignant tissues by immunohistology. As shown
here, each of the currently available CD97 mAbs binding the
shortest isoform detects 1 out of 2 immunodominant epitopes
formed by the first EGF domain (CD97EGFmAbs) or located
within the stalk region (CD97stalkmAbs). CD97EGFmAbs showed
a more restricted staining pattern compared to CD97stalkmAbs.
Differences are most conspicuous in smooth muscle cells that
strongly express the CD97stalkbut not the CD97EGFepitope.
Thus, the second aim of this study was to determine the under-
lying reason(s) for this difference. We observed that cell lines from
colon carcinomas and smooth muscle cells expressed CD97 with
different molecular weights, suggesting a cell type-specific post-
translational modification. Glycosidase digestion of CD97, expres-
sion of wild-type (wt) CD97 cDNA in glycosylation-deficient
CHO cells, in vitro inhibition of glycosylation and site-directed
mutagenesis were used to show that N-glycosylation within the
EGF domains is essential for CD97EGFmAb and CD55 ligand
MATERIAL AND METHODS
BL-Ac/F2,14VIM3b15and CLB-CD97/19bind to the first EGF
domain and are thus named CD97EGFmAbs. Unknown at the
moment of description, they also bind EMR2.16EMR2 is highly
restricted to myeloid cells7,17,18and is only rarely found in tumor
Grant sponsor: the German Research Council; Grant number: AU 132/
3-1; Grant sponsor: the Netherlands Organization for Scientific Research;
Grant number: 0409-18-02; Grant sponsor: the Interdisciplinary Center for
Clinical Research Leipzig at the Faculty of Medicine, University of
Leipzig; Grant number: D6.
Dr. Hamann is a fellow of the Royal Netherlands Academy of Arts and
*Correspondence to: Institute of Anatomy, University of Leipzig, Ph.-
Rosenthal-Strasse 55, Leipzig, 04103, Germany. Fax: ?49-341-972-2029.
Received 15 April 2004; Accepted after revision 27 May 2004
Published online 15 July 2004 in Wiley InterScience (www.interscience.
Int. J. Cancer: 112, 815–822 (2004)
© 2004 Wiley-Liss, Inc.
Publication of the International Union Against Cancer
cell lines and tissues.7,18An EMR2-specific mAb17was used as a
control to ensure CD97 specificity (data not shown). EMR2 does
not interact with CD55.16MEM-180 (kindly provided by V.
Horecs ˇı ´, University of Prague) and CLB-CD97/319bind to the
stalk region (CD97stalkmAbs) and do not show any crossreactivity
with EMR2. The CD97 mAbs used in this study are described in
CD97 immunostaining and scoring of tissues
The study was approved by the local committee of medical
ethics; all patients gave their written consent. The characteristics of
the patients with colorectal (n ? 81), gastric (n ? 50) and pan-
creatic (n ? 17) carcinomas have been published previously.6,7
Serial frozen sections of tumors and the corresponding normal
tissues were incubated using CD97EGF(clone BL-Ac/F2) and
CD97stalk(clone CLB-CD97/3) mAbs (4°C, overnight). Bound
mAbs were detected with a supersensitive detection kit (DakoCy-
tomation, Hamburg, Germany). Immunoreactivity was scored
semiquantitatively as described.6
Normal cells and tumor cell lines
Colon smooth muscle cells (BioWhittaker Europe, Verviers,
Belgium) were cultured in smooth muscle cell basal medium.
Peripheral blood mononuclear cells were isolated by Ficoll density
gradient centrifugation and stimulated with phorbol-myristate ac-
etate (PMA; 10 ng/ml; Sigma-Aldrich Chemie, Taufkirchen, Ger-
many) for 4 hr. Human colorectal (n ? 18), gastric (n ? 4) and
pancreatic (n ? 3) tumor cell lines have been described else-
where.6,7COS and CHO cells and CHO-derived glycosylation-
deficient clones Lec1 and Lec2 were obtained from the American
Type Culture Collection (ATCC, Rockville, MD).
Cells were stained with mAbs by an indirect immunofluores-
cence method using an F(ab?)2fragment of FITC-labeled or phy-
coerythrin-labeled goat antimouse Ig (FITC-GAM, PE-GAM;
Dako, Glostrup, Denmark) and analyzed by FACSscan (Becton
Dickinson, Mountain View, CA). To determine the level of CD97,
mean fluorescence intensity was analyzed and compared to control
cells stained with an irrelevant isotype-matched mAb. The com-
parison between the groups was performed with the Mann-Whit-
ney test. The Spearman test was used for correlation analysis.
CD97 mAbs MEM-180, CLB-CD97/3, CLB-CD97/1 and BL-
(Sigma). The VIM3b CD97 mAb was only used to block binding;
5 ? 105Colo 205 cells were incubated in saturated concentrations
(200 nM) of unlabeled mAbs for 30 min on ice. Cells were then
centrifuged, the supernatant was removed and the cell pellet was
incubated with the competing biotinylated mAb (60 nM) for an
additional 30 min on ice. After washing, the cells were stained
with streptavidin-PE (SA-PE; Dako) and analyzed by flow cytom-
Titrating CD97EGFand CD97stalkmAb reactivity
mAb concentration was determined using the Easy Titer Mouse
IgG kit and purified mouse IgG as a standard (Perbio Science,
Bonn, Germany). About 5 ? 105activated lymphocytes were
independently incubated with various concentrations of each mAb
type (2–1,000 nM) for 30 min on ice followed by FITC-GAM and
analyzed by flow cytometry.
Binding of CD55 to CD97 was assayed on CD97 cDNA-
transfected COS cells incubated with human erythrocytes as de-
Biochemical characterization of CD97
Surface labeling of cells was performed using D-Biotin-N-
hydroxysuccinimide ester (Roche Diagnostics, Mannheim, Germa-
ny).21Immunoprecipitation was carried out with 10 ?g mAb
CLB-CD97/3 and 40 ?l protein G plus/protein A agarose (Calbio-
chem-Novabiochem, Schwalbach, Germany). Samples were elec-
trophoresed through 7.5% SDS polyacrylamide gel and transferred
to nitrocellulose, which was probed with streptavidin-alkaline
phosphatase (Dako) and processed with nitroblue tetrazolium/5-
bromo-4-chloro-3-indolyl phosphatase (NBT/BCIP; Roche). Deg-
lycosylation reactions were conducted straight after immunopre-
cipitation with 10 mU sialidase, 2 mU O-glycosidase, or 5 U
N-glycosidase F (Roche) for 5 hr at 37°C. In order to confirm
enzyme reactivity, we digested immunoprecipitates precipitated
with CD44 mAb Hermes-3 in parallel (not shown). CD44 is known
to be highly O- and N-glycosylated.
CD97 glycosylation analysis within cell culture
The results from glycosidase digestions were confirmed by
culturing Colo 205 and smooth muscle cells with 0.1 mU/ml
sialidase (Roche) for 1 hr to hydrolyze the chain-terminating sialic
acid residues, 5 ?g/ml tunicamycin (Sigma) for 24 hr to inhibit
N-glycosylation, or 2 mM benzyl 2-acetamido-2deoxy-?-D-galac-
topyranoside (GalNAc?-O-bn; Sigma) for 72 hr to inhibit O-
glycosylation.22 After treatment, binding of CD97EGF
CD97stalkmAbs was measured by flow cytometry. The molecular
weight of the remaining CD97 was examined by Western blotting.
Furthermore, CHO cells and the Lec1 and Lec2 mutant CHO
clones were transiently transfected with human CD97 (EGF 1,2,5)
cDNA in pcDNA3.1/Zeo(?)10using lipofectamin. CD97EGFand
CD97stalkmAb binding was measured after 48 hr by flow cytom-
Mutation of N-glycosylation sites
CD97 (EGF 1,2,5) contains 8 potential N-glycosylation sites.
The first 2 sites are located within the first EGF domain, the third
site within the second EGF domain. The other 5 sites are located
within the stalk. Site-directed mutants were generated by PCR
using overlapping mutated primers, the Quick Change XL Site
Mutagenesis Kit (Stratagene Europe, Amsterdam, The Nether-
lands) and CD97 (EGF 1,2,3) cDNA in pcDNA3.1/Zeo(?)23as
template. COS cells were transiently transfected with the mutated
CD97 cDNA constructs to produce CD97 molecules in which
asparagine residues (N) in potential N-glycosylation sites were
substituted with isoleucine (I). Protein expression was verified by
staining with CD97EGFor CD97stalkmAbs after 48 hr by flow
TABLE I – CD97 MONOCLONAL ANTIBODY CROSSCOMPETITION IN BINDING OF CD97 AT THE CELL SURFACE OF COLO 205 CELLS
Binding region at CD97References
First EGF domain
First EGF domain
First EGF domain
Values indicate the percentage decrease in binding for competing mAbs obtained in the presence of saturating concentrations of blocking
mAbs. Absence of binding inhibition is indicated by ?.
WOBUS ET AL.
CD97 mAbs recognize 2 distinct epitopes
Competition experiments were carried out in which each of the
CD97 mAbs used competed with itself and every other mAb
(Table I). Efficient competition was observed between those mAbs
that recognize the same epitope. Although VIM3b also binds
within the first EGF domain,10it did not completely block BL-
Ac/F2 or CLB-CD97/1 binding. mAbs binding to the first EGF
domain had no effect on mAbs binding to the stalk region, or vice
CD97EGFand CD97stalkmAbs have similar binding affinities to
their epitopes. Titrating mAb binding on activated lymphocytes
revealed no significant divergence between the cell-surface bind-
ing profiles of the mAbs at a concentration of 60 nM or higher
(Fig. 1a), which has been used for further studies.
We next compared CD97EGFand CD97stalkmAb binding to
colorectal, gastric and pancreatic tumor cell lines. These cell lines
expressed CD97 with varying intensity.6,7There was a strong
correlation between CD97EGFand CD97stalkmAb staining inten-
sity (Fig. 1b). This indicates that the epitopes on in vitro cultured
tumor cell lines are freely accessible for CD97EGFand CD97stalk
mAbs, and that both groups of mAbs detect the same antigen. In
contrast, cultured smooth muscle cells from the third passage were
only faintly stained with CD97EGFbut strongly with CD97stalk
mAbs (Fig. 1c).
Expression of CD97 epitopes in normal tissues and carcinomas
Smooth muscle cells of the lamina muscularis mucosae and the
muscular coat throughout the gastrointestinal tract were strongly
stained with CD97stalkmAbs, whereas CD97EGFmAbs showed
only faint staining (Fig. 2a and b). Normal pancreas was
CD97EGF-negative (Fig. 2d). In contrast, ducts outlined by flat-
tened or cuboidal epithelium expressed the CD97stalkepitope
Immunohistologic analysis of colorectal (Fig. 2e–h), gastric and
pancreatic carcinomas (not shown) with both groups of CD97
mAbs revealed a completely different staining pattern. Carcinomas
frequently expressed CD97stalkbut not CD97EGFwithin tumor
cells (Fig. 2e and f). Moreover, scattered tumor cells were more
heavily stained by CD97stalkbut not CD97EGFmAbs than cells
located within solid tumor formations of the same tumor (Fig. 2g
and h). The number of positive tumors and, in some cases, the
labeling score were lower for the CD97EGFepitope than for the
CD97stalkepitope in various carcinomas (Table II). In contrast to
the results obtained with a CD97stalkmAb,6comparison of clini-
copathologic features between CD97EGF-positive and CD97EGF-
negative colorectal carcinomas revealed no differences. Clinical
grade and invasiveness are not correlated with global downregu-
lation of glycosylation (not shown). In summary, the selection of
CD97 mAbs significantly influences the prognostic value of CD97
in clinical studies.
Cell type-specific CD97 N-glycosylation
As smooth muscle cells showed the clearest difference in the
staining intensities with both groups of CD97 mAbs in vitro and in
situ, we used this cell type in comparison to the colorectal tumor
cell line Colo 205 in further experiments. No CD97 isoform-
specific mAbs that work in Western blotting are available.
Two bands of 78 and 83 kDa were detected after immunopre-
cipitation of Colo 205 cells (Fig. 3, lane 1). The same signals were
obtained in DLD-1 and LS174T cells (not shown). In contrast, 2
bands of 52 and 57 kDA were present in smooth muscle cells (Fig.
3, lane 4). This cell type-specific difference in molecular weight
may reflect a modification in posttranslational core-protein pro-
cessing. The appearance of only 2 bands suggests that one isoform,
most likely CD97 (EGF 1,2,3,4,5), is not present in either tumor or
smooth muscle cells. We confirmed this assumption with an mAb
against the 4th EGF domain, which is present only in the longest
CD97 isoform. This mAb did not stain tumor or smooth muscle
cells in flow cytometry (not shown).
To characterize the glycan structures, Colo 205, smooth muscle
cells and CD97 immunoprecipitates of these cells were treated
with sialidase, which hydrolyzes chain-terminating sialic acid res-
idues. No influence on CD97EGFmAb binding or CD97 molecular
size was observed (not shown). In addition, CD97EGFmAb bind-
ing was not affected in CD97-transfected Lec2 cells that exhibit a
drastic reduction in the transport of CMP-sialic acid into the Golgi
compartment (not shown).
CD97 contains several potential O-glycosylation sites. There-
fore, we digested CD97 sialidase-treated immunoprecipitates with
O-glycosidase. The molecular size of CD97 was not changed in
either cell type (Fig. 3, lanes 2 and 5). To confirm this result, the
cells were cultured with an O-glycosidase inhibitor. No shift in the
size of CD97 protein could be detected. CD97EGFand CD97stalk
mAb binding was not influenced (not shown).
Finally, CD97 immunoprecipitates were treated with N-glyco-
sidase F, which revealed the protein free from asparagine-linked
carbohydrates. CD97 was deglycosylated to 52 and 57 kDa in Colo
205 cells (Fig. 3, lane 3). The same treatment did not affect the size
of CD97 immunoprecipitated from smooth muscle cells (lane 6),
indicating that the 52 and 57 kDa products probably represent 2
unglycosylated isoforms of CD97. Together, the results indicate
the existence of cell type-specific N-glycosylation of CD97.
N-glycosylation within EGF domains is important for CD97EGF
Ineffective N-glycosylation may cause marginal CD97EGFmAb
binding in colon smooth muscle cells. To test this hypothesis, we
transfected Lec1 cells with CD97 cDNA. The cells lack GlcNAc
glycosyl transferase, so N-linked carbohydrates are blocked at the
Man5-GlcNAC2-Asn intermediate. However, this N-glycosylation
defect did not influence CD97 mAb binding (not shown).
For complete inhibition of N-type glycosylation, Colo 205 cells
were treated with tunicamycin. After 24 hr, immunoprecipitated
CD97 was found to present the deglycosylated 52–57 kDa forms in
FIGURE 1 – Reactivity of CD97 mAbs on different cell types. (a)
CD97EGFand CD97stalkmAb binding on the cell surface of activated
peripheral blood lymphocytes. The data points represent the means of
3 independent experiments (mean ? SEM). (b) Correlation between
CD97EGFand CD97stalkexpression levels on colorectal, gastric and
pancreatic carcinoma cell lines (n ? 25) determined as mean fluores-
cence intensity in flow cytometry. The data points represent the means
of 3 independent experiments. (c) Staining of cultured colon smooth
muscle cells in the third passage with CD97stalkand CD97EGFmAbs.
N-GLYCOSYLATION OF CD97
some cases (Fig. 4b, lane 3). The remaining fraction of CD97
migrated as the glycosylated 75–85 kDa forms. The 52–57 kDa
bands were similar to those of N-glycosidase F-digested CD97
(Fig. 4b, lane 2). CD97 present on the cell surface during tunica-
mycin treatment remained N-glycosylated, whereas newly synthe-
sized CD97 was free of N-glycans. Longer tunicamycin treatment
to increase the N-glycan-free CD97 fraction was not recommended
due to this substance’s severe side effects.
Importantly, tunicamycin significantly decreased CD97EGF
mAb binding compared to untreated control cells in a time-depen-
dent manner as shown by mean fluorescence intensity and number
of positive cells in flow cytometry (Fig. 4a). Due to the effect of
tunicamycin on cell viability, CD97stalkmAb staining intensity
also slightly decreased but the number of CD97stalk-stained cells
remained unaffected. The results indicate that CD97EGFmAb
binding depends on N-glycosylation.
To determine which N-glycosylation sites of CD97 (EGF 1,2,5)
are crucial for mAb binding, individual potential N-linked glyco-
sylation site-specific mutants were generated. Expression levels of
the mutants as depicted by staining with CD97stalkmAbs in flow
cytometry was comparable to wt CD97 (Fig. 5), indicating that all
mutants were correspondingly expressed on the surface. Interest-
ingly, mutation of the second N-linked glycosylation site at posi-
FIGURE 2 – Immunohistologic expression pat-
tern of CD97 epitopes in serial sections of nor-
mal and malignant tissues. (a) and (b) Normal
colon. Smooth muscle cells of the lamina mus-
cularis mucosae and the muscular coat (arrows)
were strongly positive for CD97stalkbut nearly
CD97EGF-negative. A leukocyte subpopulation
expressed CD97EGF(arrow head). (c) and (d)
Normal pancreas. Exocrine pancreatic cells
showed a faint staining for CD97stalk, whereas
the cells of pancreatic ducts (open arrow) ex-
pressed the molecule. In contrast, CD97EGF
mAbs did not stain pancreatic structures. (e) and
(f) Colon carcinoma. Smooth muscle cells (ar-
row) were CD97stalk-positive but CD97EGF-neg-
ative. Tumor cells expressed CD97stalkbut not
CD97EGF(open arrow). Lymphocytes located
within a follicle (upper part of the picture) did
not express CD97, whereas a few leukocytes
located in the stroma were CD97EGF-positive
(arrow head). (g) and (h) Rectum carcinoma.
Stronger expression of CD97stalkin scattered
tumor cell groups surrounded by stroma (open
arrow) compared to tumor cells located in tumor
glands or solid tumor trabecula. Homogenous
staining of tumor cells by CD97EGFmAbs (open
arrow). Smooth muscle cells express CD97stalk
CD97EGF(arrowhead). Scale bars ? 50 ?m.
TABLE II – NUMBER OF CD97EGF-POSITIVE AND CD97stalk-POSITIVE
CARCINOMAS IN IMMUNOHISTOLOGY
3.3 ? 0.5
4.1 ? 2.7
6.0 ? 2.0
5.7 ? 2.41
7.1 ? 3.6
6.1 ? 2.340
The score6describes the product of positive cells and staining
intensity (0–12). Samples were positive at a score higher than 2. A
score of 3–6 was regarded as moderate expression; a score over 6 as
strong expression.–1p ? 0.01.
WOBUS ET AL.
tion 38 located within the first EGF domain resulted in an almost
complete abrogation of CD97EGFmAb binding (3.2% ? 1.2%).
The 2 other potential N-glycosylation sites within the EGF do-
mains also contributed to this phenomenon. Mutation of the first
N-glycosylation site (N33I) decreased CD97EGFmAb binding to
23.1% ? 7.3% compared to wt CD97. The third N-glycosylation
site at position 108 only weakly contributes to CD97EGFepitope
accessibility (80.7% ? 8.6%). The role of the first and third
N-glycosylation site became more evident when both sites were
mutated in parallel (N33I ? N108I; 8.5% ? 3.3%). Mutation of
N-glycosylation sites within the stalk region had no influence on
either CD97EGFor CD97stalkmAb binding.
N-glycosylation of EGF domains influences CD55 binding
The binding site for CD55 is located within the first 2 CD97
EGF domains.10,11To assess whether glycosylation of the EGF
domains has an effect on CD55 binding, wt CD97 cDNA-trans-
fected Lec1 and Lec2 cells and N-glycan CD97 cDNA mutant-
transfected COS cells were used to quantify the adhesion of
erythrocytes expressing CD55. Erythrocytes bound to COS and to
glycosylation-deficient Lec1 or Lec2 cells expressing wt CD97 at
Interestingly, all mutants lacking the glycosylation at position
38 almost failed to bind erythrocytes (Fig. 6). The binding was also
reduced to COS cells expressing either the single N33I and N108I
mutants or a combination of both. The result demonstrates that
N-glycosylation within the EGF domains is important for CD55
ligand binding. In contrast, N-glycosylation mutants in the stalk
region showed no difference to wt CD97-transfected control cells
The immunodominant epitopes of CD97 are not equally present
on normal and malignant cells and tissues. The detection of dif-
ferent molecular weights but similar mRNA splicing pattern6in
smooth muscle cells and tumor cell lines suggested a cell type-
specific posttranslational modification of CD97. Examining the
glycosylation pattern, we demonstrated that N-linked glycans may
be a part of the CD97EGFmAb-binding site, or are required for its
appropriate conformation. Defective N-glycosylation reduces the
molecular weight of CD97 in smooth muscle cells to 52–57 kDa,
corresponding to the nonglycosylated protein core,24and, in par-
allel, impairs CD97EGFmAb binding.
N-glycosylation of epitopes plays an important role in the rec-
ognition of determinants, folding of proteins and establishment of
their biologic activity.25–27Cell type-specific posttranslational
modifications of EGF domains28may provide a general mecha-
nism for regulating ligand-receptor interactions. Aspartyl beta-
hydroxylation and O-fucose glycosylation of EGF domains mod-
ulate Notch-receptor signaling.29,30Despite the unusually high
serine/threonine content, O-glycosylation seems to be less relevant
to CD97 molecular weight and interactions. Neither O-glycosidase
treatment of immunoprecipitates nor inhibition of O-glycosylation
in culture influenced CD97 molecular weight or mAb and CD55
binding. Extensive O-linked glycosylation of the stalk region has
been predicted for EGF-TM7 molecules as reflected by the desig-
nation EGF module-containing mucin-like receptor (EMR) in sev-
eral family members.31However, in agreement with our results on
CD97, detailed biochemical analysis of mouse EMR1 provided no
evidence for substantial O-linked glycosylation.32
In contrast to smooth muscle and carcinoma cells in situ, acces-
sibility of CD97stalkand CD97EGFepitopes on tumor cell lines
FIGURE 3 – Biochemical characterization of CD97 in Colo 205 and
colon smooth muscle cells. Immunoprecipitation with CD97stalkmAbs
revealed a 78 and 83 kDa signal in Colo 205 cells representing the
common glycosylated forms of CD97 (lane 1), whereas a 52 and 57
kDa band was present in smooth muscle cells (lane 4). Treatment with
O-glycosidase had no influence on the molecular weight in both cell
types (lanes 2, 5). N-glycosidase F digestion reduced the molecular
weight of CD97 in Colo 205 cells to 52–57 kDa (lane 3) but did not
influence that in smooth muscle cells (lane 6).
FIGURE 4 – Effect of the inhibition of N-glycosylation by tunicamy-
cin in Colo 205 cells. (a) Flow cytometric analysis of CD97EGFand
CD97stalkmAb binding after treatment of cells with tunicamycin for
12, 24 and 36 hr. Mean fluorescence intensity (upper number) and
percentage of positive cells (lower number) are indicated. (b) Inhibi-
tion of N-glycosylation caused a reduction of the CD97 molecular
weight in newly synthesized molecules to approximately 52–60 kDa
(lane 3). A similar reduction in size was achieved by treatment of
CD97 immunoprecipitates with N-glycosidase F (lane 2).
N-GLYCOSYLATION OF CD97
correlates well. The discrepancy between tumor cells in solid
tissue and nonsolid structures suggested that besides N-glycosyl-
ation of the EGF domains, other mechanisms may affect epitope
accessibility. One of our hypotheses was that a ligand to CD97
sterically hinders mAb binding in tumor tissues, whereas this
ligand may not be present or replaced from the binding site in
nonsolid formations. Colorectal tumors, for example, highly ex-
press the CD97 ligand CD55 in the stroma.33CD55 may be shed
from the cells, bind to CD97 and mask the CD97EGFepitope.
However, incubating CD97-transfected COS cells with the ligand
CD55 did not impair CD97EGFmAb binding and differential
proteolysis or microwave antigen retrieval techniques to unmask
the antigen were not successful (data not shown). Finally, macro-
phages in the intimal lining layer in rheumatoid synovium are
strongly stained by CD97EGFmAbs, although neighboring fibro-
blast-like synoviocytes express high CD55 levels.34
We also excluded N-terminal protein truncation due to muta-
tions or alternative mRNA splicing and destabilization of EGF
domains through Ca2?inaccessibility10as mechanisms that might
impair binding of CD97EGFmAbs to smooth muscle cells. In
addition, CD97stalkmAbs did not crossreact with the related EGF-
TM7 receptor ETL35expressed by smooth muscle (data not
N-glycosylation alters function as shown convincingly for a
number of glycoproteins.26,27,36Thus, incomplete or missing N-
glycosylation of the EGF domains in CD97stalk-positive but
CD97EGF-negative or slightly positive cell types might have func-
tional consequences. As shown here, N-glycosylation of CD97
within the first 2 EGF domains, which mediates the binding to
CD55,10,11is required for this interaction. Consequently, it can be
predicted that CD97 on smooth muscle cells has little if any
affinity for CD55. Scattered tumor cells located at the invasion
front of colorectal carcinomas known to invade and metasta-
size37,38showed an overexpression of CD97 detected with
CD97stalkmAbs but not CD97EGFmAbs. Besides the possibility
that the CD97EGFepitope is covered by a ligand in situ, our data
suggest that these cells overexpress partly deglycosylated CD97.
Whether the tumor cells use CD55 or another ligand for interaction
to leave the well-formed solid structures and to invade into the
extracellular matrix via CD97 remains unclear.
Interestingly, ligand-binding properties of CD97 molecules,
which lack the CD97EGFepitope, resemble EMR2. EMR2 is an
EGF-TM7 receptor expressed on myeloid cells, the EGF domain
region of which is 97% identical to that of CD97.17,39Other than
CD97, EMR2 does not recognize CD55.12,39Three different amino
acids within the first 2 EGF domains account for this difference in
ligand binding.11The first different amino acid is Asn33in CD97,
one of the N-glycosylation sites within the first EGF domain of
CD97. The difference in CD55 ligand specificity between CD97
and EMR2 is likely to be at least partly due to different N-
Taken together, our results show that there are CD97 protein
modifications that mask or uncover epitopes for certain CD97
mAbs. The impact of N-glycosylation on accessibility of the
CD97EGFepitope has consequences for the value of CD97 in
pathology. The results are relevant with regard to several aspects
of previous and future studies.
First, the choice of the mAb strongly influences immunohisto-
logic data on the correlation between CD97 and histopathologic
subtypes, diagnosis, progression and prognosis. For example, the
CD97EGFbut not the CD97stalkepitope, in combination with CD95
FIGURE 5 – The importance of N-glycosylation within the EGF domains in CD97EGFmAb binding. (a) Schematic presentation of the shortest
CD97 (EGF 1,2,5) isoform. The potential N-glycosylation sites are indicated. (b) Flow cytometric analysis of the binding of CD97EGFand
CD97stalkmAbs to COS cells transiently transfected with wt CD97 cDNA (EGF 1,2,5) or CD97 N-linked glycosylation site-specific mutants.
(c) Percentage of binding of CD97EGFmAbs in COS cells transiently transfected with wt CD97 cDNA (EGF 1,2,5) or CD97 N-linked
glycosylation site-specific mutants (n ? 4; mean ? SEM).
WOBUS ET AL.
and Fas-L, is a useful marker for distinguishing pancreatitis
from pancreatic duct cell carcinoma.13Both CD97 epitopes
need to be carefully compared in immunohistologic studies for
a correct description of CD97 expression. Second, previous
negative findings on soluble CD97 in sera of colorectal cancer
patients need to be evaluated.6These results are based on a
CD97stalkmAb for coating and a CD97EGFmAb for detection.34
As we know now, the assay could only detect N-glycosylated
soluble CD97. Third, we examined the results of our own study
that focused on the distribution of CD97 on leukocytes, espe-
cially in lymphoid tissues.3Results published in this study
corresponded well to those obtained with CD97EGFbut not
CD97stalkmAbs. Fourth, cell type-specific binding of CD97
epitope-specific mAbs is of importance in studies targeting
CD97-positive cells. CD97stalkmAbs would not only bind tu-
mor cells, but also smooth muscle cells.
In summary, CD97 mAbs used for a particular subject must be
chosen with the specific goal of the study in mind, since the results
vary significantly and the data can be interpreted only if the mAb
epitope specificity is known. Our data, coupled with the observa-
tion that N-glycosylation alters CD55 ligand binding to CD97,
provide a mechanism for cell type-specific diversity of both struc-
ture and function of EGF-TM7 molecules.
The authors thank M. Steinert (Department of Surgery), A.
Schu ¨tz (Institute of Pathology, University of Leipzig) and C.
Boltze (Institute of Pathology, University of Magdeburg) for pro-
viding the tissues; M. Wahlbuhl, D. Sittig and S. Gawlowska for
their contribution in performing experiments; M. Kwakkenbos
(University of Amsterdam) for critical examination of the manu-
script; and A. Garritsen and A. van Elsas (NV Organon, Oss, The
Netherlands) for kindly providing soluble CD55.
1.McKnight AJ, Gordon S. The EGF-TM7 family: unusual structures at
the leukocyte surface. J Leukoc Biol 1998;63:271–80.
Kwakkenbos MJ, Kop EN, Stacey M, Matmati M, Gordon S, Lin HH,
Hamann J. The human EGF-TM7 family: a postgenomic view. Im-
Jaspars LH, Vos W, Aust G, van Lier RA, Hamann J. Tissue distri-
bution of the human CD97 EGF-TM7 receptor. Tissue Antigens
Aust G, Eichler W, Laue S, Lehmann I, Heldin N-E, Lotz O,
Scherbaum WA, Dralle H, Hoang-Vu C. CD97: a dedifferentiation
marker in human thyroid carcinomas. Cancer Res 1997;57:1798–806.
Hoang-Vu C, Bull K, Schwarz I, Krause G, Schmutzler C, Aust G,
K¨ ohrle J, Dralle H. Regulation of CD97 protein in thyroid carcinoma.
J Clin Endocrinol Metab 1999;84:1104–9.
Steinert M, Wobus M, Boltze C, Sch¨ utz A, Wahlbuhl M, Hamann J,
Aust G. Expression and regulation of CD97 in colorectal carcinoma
cell lines and tumor tissues. Am J Pathol 2002;161:1657–67.
Aust G, Steinert M, Sch¨ utz A, Wahlbuhl M, Hamann J, Wobus M.
CD97, but not its closely related EGF-TM7 family member EMR2, is
expressed on gastric, pancreatic and esophageal carcinomas. Am J
Clin Pathol 2002;118:699–707.
Stacey M, Lin HH, Gordon S, McKnight AJ. LNB-TM7, a group of
seven-transmembrane proteins related to family-B G-protein-coupled
receptors. Trends Biochem Sci 2000;25:284–9.
Hamann J, Vogel B, van Schijndel GM, van Lier RA. The seven-span
transmembrane receptor CD97 has a cellular ligand (CD55, DAF). J
Exp Med 1996;184:1185–9.
10. Hamann J, Stortelers C, Kiss-Toth E, Vogel B, Eichler W, van Lier
RA. Characterization of the CD55 (DAF)-binding site on the seven-
span transmembrane receptor CD97. Eur J Immunol 1998;28:1701–7.
11. Lin HH, Stacey M, Saxby C, Knott V, Chaudhry Y, Evans D, Gordon
S, McKnight AJ, Handford P, Lea S. Molecular analysis of the
epidermal growth factor-like short consensus repeat domain-mediated
protein-protein interactions: dissection of the CD97-CD55 complex.
J Biol Chem 2001;276:24160–9.
12. Stacey M, Chang GW, Davies JQ, Kwakkenbos MJ, Sanderson RD,
Hamann J, Gordon S, Lin HH. The epidermal growth factor-like
domains of the human EMR2 receptor mediate cell attachment
through chondroitin sulphate glycosaminoglycans. Blood 2003;102:
13. Boltze C, Schneider-Stock R, Aust G, Mawrin C, Dralle H, Roessner
A, Hoang-Vu C. CD97, CD95 and Fas-L clearly discriminate between
chronic pancreatitis and pancreatic ductal adenocarcinoma in periop-
erative evaluation of cryocut sections. Pathol Int 2002;52:83–8.
14. Eichler W, Aust G, Hamann D. Characterization of the early activa-
tion-dependent antigen on lymphocytes defined by the monoclonal
antibody BL-Ac(F2). Scand J Immunol 1994;39:111–5.
15. Pickl WF, Majdic O, Mai I, Gadd S, Knapp W. Overview of CD97.
In: Schlossman SF, Boumsell L, Gilks W, Halan JM, Morimoto C,
Ritz J, Shaw S, Silverstein R, Springer T, Tedder TF, Todd RF, eds.
Leucocyte typing V. Oxford: Oxford University Press, 1995. 1151–3.
16. Lin HH, Stacey M, Hamann J, Gordon S, McKnight AJ. Human
EMR2, a novel EGF-TM7 molecule on chromosome 19p13.1, is
closely related to CD97. Genomics 2000;67:188–200.
17. Kwakkenbos MJ, Chang GW, Lin HH, Pouwels W, de Jong EC, van
Lier RA, Gordon S, Hamann J. The human EGF-TM7 family member
EMR2 is a heterodimeric receptor expressed on myeloid cells. J Leu-
koc Biol 2002;71:854–62.
18. Aust G, Hamann J, Schilling N, Wobus M. Detection of alternatively
spliced EMR2 mRNAs in colorectal tumor cell lines but rare expres-
sion of the molecule in the corresponding adenocarcinomas. Virch
19. Kwakkenbos MJ, van Lier RA, Hamann J. Characterization of EGF-
TM7 family members by novel monoclonal antibodies. In: Mason D,
ed. Leucocyte typing VII: white cell differentiation antigens. Oxford:
Oxford University Press, 2002. 381–3.
20. Qian YM, Haino M, Kelly K, Song WC. Structural characterization of
mouse CD97 and study of its specific interaction with the murine
decay-accelerating factor (DAF, CD55). Immunology 1999;98:303–
FIGURE 6 – Binding of human erythrocytes to COS cells transiently
transfected with wt CD97 cDNA (EGF 1,2,5) or CD97 N-linked
glycosylation site-specific mutants. Erythrocytes express CD55, a li-
gand of CD97. (a) Erythrocytes bound to COS cells transfected with
wt CD97 cDNA, whereas they failed to bind to the CD97 N38I mutant.
(b) The erythrocyte binding capacity was compared between nontrans-
fected COS cells, COS cells transfected with wt CD97 cDNA and the
CD97 N-linked glycosylation mutants by measuring the hemoglobin
content of lysed adhered erythrocytes (n ? 4; mean ? SEM). Asterisk,
p ? 0.01 compared to wt CD97 cDNA-transfected cells.
N-GLYCOSYLATION OF CD97
21. Cole SR, Ashman LK, Ey PL. Biotinylation: an alternative to radio- Download full-text
iodination for the identification of cell surface antigens in immuno-
precipitates. Mol Immunol 1987;24:699–705.
22. Gouyer V, Leteurtre E, Zanetta JP, Lesuffleur T, Delannoy P, Huet G.
Inhibition of the glycosylation and alteration in the intracellular traf-
ficking of mucins and other glycoproteins by GalNAcalpha-O-bn in
mucosal cell lines: an effect mediated through the intracellular syn-
thesis of complex GalNAcalpha-O-bn oligosaccharides. Front Biosci
23. Hamann J, Eichler W, Hamann D, Kerstens HM, Poddighe PJ,
Hoovers JM, Hartmann E, Strauss M, van Lier RA. Expression
cloning and chromosomal mapping of the leukocyte activation antigen
CD97, a new seven-span transmembrane molecule of the secretion
receptor superfamily with an unusual extracellular domain. J Immunol
24. Gray JX, Haino M, Roth MJ, Maguire JE, Jensen PN, Yarme A,
Stetler-Stevenson MA, Siebenlist U, Kelly K. CD97 is a processed,
seven-transmembrane, heterodimeric receptor associated with inflam-
mation. J Immunol 1996;157:5438–47.
25. Samih N, Hovsepian S, Notel F, Prorok M, Zattara-Cannoni H,
Mathieu S, Lombardo D, Fayet G, El Battari A. The impact of N- and
O-glycosylation on the functions of Glut-1 transporter in human
thyroid anaplastic cells. Biochim Biophys Acta 2003;1621:92–101.
26. Schroter S, Derr P, Conradt HS, Nimtz M, Hale G, Kirchhoff C.
Male-specific modification of human CD52. J Biol Chem 1999;274:
27. Edberg JC, Kimberly RP. Cell type-specific glycoforms of Fc gamma
RIIIa (CD16): differential ligand binding. J Immunol 1997;159:3849–
28. Harris RJ, Spellman MW. O-linked fucose and other post-transla-
tional modifications unique to EGF modules. Glycobiology 1993;3:
29. Bruckner K, Perez L, Clausen H, Cohen S. Glycosyltransferase ac-
tivity of fringe modulates notch-delta interactions. Nature 2000;406:
30. Dinchuk JE, Focht RJ, Kelley JA, Henderson NL, Zolotarjova NI,
Wynn R, Neff NT, Link J, Huber RM, Burn TC, Rupar MJ, Cun-
ningham MR, et al. Absence of post-translational aspartyl beta-hy-
droxylation of epidermal growth factor domains in mice leads to
developmental defects and an increased incidence of intestinal neo-
plasia. J Biol Chem 2002;277:12970–7.
31. Baud V, Chissoe SL, Viegas-Pequignot E, Diriong S, N’Guyen VC,
Roe BA, Lipinski M. EMR1, an unusual member in the family of
hormone receptors with seven transmembrane segments. Genomics
32. Haidl ID, Jefferies WA. The macrophage cell surface glycoprotein
F4/80 is a highly glycosylated proteoglycan. Eur J Immunol 1996;26:
33. Li L, Spendlove I, Morgan J, Durrant LG. CD55 is over-expressed in
the tumour environment. Br J Cancer 2001;84:80–6.
34. Hamann J, Wishaupt JO, van Lier RA, Smeets TJ, Breedveld FC, Tak
PP. Expression of the activation antigen CD97 and its ligand CD55 in
rheumatoid synovial tissue. Arthritis Rheum 1999;42:650–8.
35. de Vries JE, Dinjens WN, De Bruyne GK, Verspaget HW, van der
Linden EP, de Bruine AP, Mareel MM, Bosman FT, ten Kate J. In
vivo and in vitro invasion in relation to phenotypic characteristics of
human colorectal carcinoma cells. Br J Cancer 1995;71:271–7.
36. Ashman LK, Aylett GW, Cambareri AC, Cole SR. Different epitopes
of the CD31 antigen identified by monoclonal antibodies: cell type-
specific patterns of expression. Tissue Antigens 1991;38:199–207.
37. Compton CC, Fielding LP, Burgart LJ, Conley B, Cooper HS, Ham-
ilton SR, Hammond ME, Henson DE, Hutter RV, Nagle RB, Nielsen
ML, Sargent DJ, et al. Prognostic factors in colorectal cancer: College
of American Pathologists consensus statement 1999. Arch Pathol Lab
38. Masaki T, Muto T. Predictive value of histology at the invasive
margin in the prognosis of early invasive colorectal carcinoma. J
39. Lin HH, Stubbs LJ, Mucenski ML. Identification and characterization
of a seven transmembrane hormone receptor using differential dis-
play. Genomics 1997;41:301–8.
40. Hamann J, Eichler W, Vogel B, Stortelers C, Hamann D, Horejsi V,
van Lier RA. CD97 workshop panel report. In: Kishimoto T, Kikutani
H, von dem Borne AEGKr, Goyert SM, Mason DY, Miyasaka M,
Moretta L, Okumura K, Shaw S, Springer TA, Sugamura K, Zola H,
eds. Leucocyte typing VI: white cell differentiation antigens. New
York: Garland Publishing, 1997. 527–9.
WOBUS ET AL.